Abstract
MRI findings are increasingly used as outcome measures in therapeutic trials in MS.The discrepancy between the extent of the lesions on conventional T2 images and the clinical condition of the patient is one of the problems encountered in such studies. This clinical-radiological paradox prevents the use of MRI data as surrogate markers of disability in MS. A recent pilot study suggested a relationship between hypointense lesions on T1 MRI and disability. To assess in more detail the correlation of changes in hypointense lesion load on T1-weighted spin-echo MR images ("black holes") with changes in disability in MS, we studied 46 patients with clinically definite MS at baseline and after a median follow-up of 40 months. There was a significant correlation between baseline disability and hypointense lesion load (Spearman rank correlation coefficient [SRCC] = 0.46, p = 0.001). In secondary progressive patients, the rate of accumulation of these "black holes" was significantly related to progression rate (SRCC = 0.81, p < 0.0001). We speculate that the appearance of hypointense lesions is the MRI equivalent of a failure of remission. Overall, T1 lesion load measurements correlated better with clinical assessments than T2 lesion load measurements. Quantification of hypointense lesion load on T (1-weighted) spin-echo MRI helps to resolve the clinical-radiological paradox between disability and MRI and has the potential to be a surrogate marker of disability in MS.
NEUROLOGY 1996;47: 1469-1476
MRI has, since its introduction in 1981, been established as the most sensitive paraclinical tool to demonstrate dissemination in space and time in MS. [1]
The publication of the results of the interferon beta-1b (IFN-beta-1b) trial and the subsequent accelerated approval of the drug by the Food and Drug Administration (partly based on a significant decrease by 75% of MRI-detected new lesion formation in the high dose treated group) has shown the potential value of MRI as a secondary outcome measure in phase III trials. [2]
On the basis of the high sensitivity of serial gadolinium-enhanced MRI for disease activity in multiple sclerosis, [3] MRI is now also used as a primary outcome measure in an increasing number of exploratory (early phase II) therapeutic trials in MS. Because serial MRI detects between 2 and 10 times more disease activity than is clinically apparent, [4] these trials can be both shorter and more conclusive, thereby speeding up the route from phase I (human toxicity) studies to phase III (clinical outcome) studies.
The appearance of new, enhancing or not, lesions is now generally accepted as the MRI equivalent of relapses. [5] Aside from the occurrence of relapses, the clinical course of MS is mainly determined by the "baseline" disability of the patient. This can be considered as the real "burden" of the disease, and the improvement or stabilization of this disability is the ultimate goal of treatment. The extent of lesions on conventional T2-weighted images is currently regarded as the corresponding MRI measure of this disease burden. A good correlation is found between the location and extent of hyperintense abnormalities seen on T2-weighted spin-echo (SE) images and the presence of pathological abnormalities of MS. [6] In cross-sectional studies, however, highly variable correlations of T2 lesion load on MRI with clinically established disability of MS patients are reported. [7,8] Some longitudinal studies have shown that changes in MRI on T2-weighted images over time correlate weakly but significantly with changes in disability. Filippi et al. found a significant correlation of 0.13 between the total number of new lesions detected visually on T2-weighted images and increase in disability. [9] In a quantitative study using computeraided lesion load measurements, Paty et al. found a significant correlation of 0.23 between increase in T2 lesion load and increase in disability. [10] Although the correlations between changes in T2 and the expanded disability status scale (EDSS) are statistically significant, clinical changes are only partially explained by MRI changes. A possible explanation for this clinical-radiological paradox, aside from methodological differences between studies in performing the MRI and in quantification of the lesions, could be that the high sensitivity of T2-weighted images for all but the initial stadium of lesion formation (edema, demyelination, remyelination, axonal loss, gliosis) precludes differentiation of those lesions causing persistent deficit. The origin of persistent deficit probably lies in the occurrence of axonal loss or extensive demyelination. [11]
Clearly, more specific MR measures for these tissue changes are needed to resolve the clinical-radiological paradox. At present, studies investigating the amount of tissue disruption using sophisticated MR techniques such as MR spectroscopy, [12] magnetization transfer imaging, [13,14] and diffusion imaging [15] have shown potential but are still under further investigation.
Standard MR imaging techniques could perhaps already provide us with some measure of persistent demyelination, axonal loss, or gliosis. Routinely, MR examinations in MS patients do not only involve T2-weighted images but also T1-weighted SE images, usually after administration of a paramagnetic contrast agent to evaluate blood-brain barrier disruption. These T1-weighted SE images often show areas of low signal intensity. Uhlenbrock et al. proposed that these hypointense lesions as seen on T1-weighted images represent areas of axonal loss and gliosis. [16] On the basis of low magnetization transfer ratios found in hypointense lesions on T1-weighted SE images, it is speculated that these hypointense lesions represent areas of substantial structural loss. [14,17]
A recent quantitative study found a significant correlation between increase in disability and increase in hypointense lesion load on T1 SE in a group of 19 patients followed for a median period of 2 years. [18] In that pilot study, the variance in clinical terms was only moderate and the results therefore only tentative. The aim of the present study was to extend these findings in a larger patient sample with a longer follow-up to confirm that standard T1-weighted imaging can indeed provide us with a partial solution of the clinical-radiological paradox in MS.
Methods.
Patients.
Forty-six patients (28 women: 18 men, median age at entry 35 years [range 21 to 53 years], median disease duration at entry 7 years [range 1 to 24 years]) were selected on the basis of having an MRI according to a standard protocol (see below) and clinical assessment performed in our centers more than 18 months before. They were then invited to be re-evaluated with the same clinical and MRI protocol. All patients had clinically definite MS according to the Poser criteria. [19] The clinical condition was evaluated at the time of the MR scans according to the EDSS (Kurtzke). [20] Because patients were examined initially in the framework of several running research projects, they were all seen by two neurologists experienced in EDSS assessment (either C.P. or O.H.). During the second visit all patients were seen by the same neurologist (L.T.).
MRI.
All scans were performed at 0.6 T with a standard head-coil (Technicare, Solon, OH). The same imaging protocol was used for the first and the second scan, with at least two pilot scans for accurate repositioning according to internal landmarks. T2-weighted SE series (TR 2755/TE 60, 120/2 excitations) and T1-weighted SE series (TR 450/TE 28, 4 excitations), after IV administration of gadopentetate acid (0.1 mmol/kg), were obtained in a double oblique axial plane. Nineteen slices with an in-plane resolution of 1.0 x 1.3 mm, and a slice thickness of 5 mm, were obtained with a gap of 1.25 mm. The line connecting the inferior border of the pituitary gland and the fastigium of the fourth ventricle was used for angulation, and the Z-center was aligned to the caudal border of the splenium of the corpus callosum. The scans were transferred to a Sparc work station (SUN, Palo Alto, CA) for quantitative analysis.
T2-weighted images were evaluated for the presence of hyperintense regions >3 mm2. T1-weighted images were evaluated for the presence of hypointense regions >3 mm2 corresponding to hyperintense regions on T2; care was taken not to include grey matter. Both T1 and T2 lesion loads were quantified by a single observer (J.V.W.), blinded to the clinical data, using a seed-growing technique with home-developed software. [18] The baseline and follow-up images were analyzed consecutively to avoid drift in operator criteria (if any) over time. To evaluate the intraobserver variation, scans of five patients with varying lesion loads were reassessed.
Statistical analysis.
All data were normalized for the duration of follow-up by dividing the change in a measurement between two observations by the follow-up period in years. This provided a score giving the change in parameter per year of observation, allowing comparison of patients with various duration of follow-up. Since most of the data were not normally distributed, medians instead of means were used to describe the data, the Spearman rank correlation coefficient (SRCC) was used for the correlations, and the Mann-Whitney U test was used for comparisons between subgroups. Because of the great number of comparisons, a two-sided significance level of <or=to0.01 was considered statistically significant, a p value between 0.01 and 0.05 as a trend, and a p value above 0.05 as not significant. Multiple regression analysis (forward and backward stepwise, F to enter = 0.5) was used to estimate the relative weight of clinical and MR measures on the changes in disability, expressed as progression rate during study (Delta EDSS per year).
Results.
Clinical and MRI follow-up.
Forty-six patients were followed up for a median period of 40 months (range 18 to 48 months). On the basis of the clinical course before entry in the study, 29 patients were considered to have relapsing-remitting MS (RRMS), with or without sequelae, but with stable disability between relapses. Seventeen patients were secondary progressive (SPMS); they showed progressive increase in disability over the 6 months before entry with or without superimposed relapses. None of the relapsing-remitting patients became secondary progressive during the study. As expected, significant clinical differences were found between the RRMS and SPMS groups (Table 1). In comparison with our previous study, [18] significant changes in disability (change in EDSS >or=to1 point) were more frequent, occurring in 13 patients (12 SPMS, 1 RRMS) during follow-up. The distribution of changes in EDSS is presented in Figure 1. In the SPMS patients, the median increase in EDSS was 2.0; in the RRMS patients this was 0. As can be deduced from Figure 1, some of the RRMS cases were first seen during or shortly after a clinical relapse, explaining the negative EDSS changes during follow-up. At the second evaluation, the shortest interval between relapse and evaluation in the RRMS group was 4 weeks in one case; for the other patients, the interval between final evaluation and a relapse was at least 3 months. During the follow-up period, the only treatments used were short courses of high dose IV steroids for relapses in both groups; no immunosuppressives were used.
Table 1. Differences between the relapsing-remitting (RRMS) and secondary progressive (SPMS) patients for clinical and imaging measures
Figure 1. Change in EDSS during follow-up for both RRMS and SPMS patients (total n = 46).
On MRIs there were identifiable differences between both groups for baseline and follow-up measurements as illustrated in Table 1. The SPMS patients tended to have higher T1 lesion loads and faster accumulation of hypointense lesions on T1. T2 lesion loads were not significantly different between RRMS and SPMS patients. For the group as a whole, a median increase in T1 lesion load of 0.46 cm2 per year was found (range 0.7 to 9.1 cm2 per year). T2 lesion load accumulation ranged from -0.5 to 13.8 cm2 per year with a median of 1.4 cm2 per year. The median ratio of changes in T1 lesion load over T2 lesion load was 0.48 (range -0.3 to 3.0).
MRI versus clinical data.
Cross-sectional comparisons: baseline and endpoint correlations.
A stronger correlation between T1 lesion load and EDSS at entry than between T2 lesion load and EDSS at entry for the group as a whole was found (Table 2). This relationship was statistically significant for T1 but only a trend for T2. The ratio of T1 to T2 lesion load also correlated significantly with the EDSS at entry (SRCC = 0.41, p = 0.005). This correlation was also found when looking at the endpoint MRI and EDSS (Table 2). Most interestingly, the correlation between the ratio of T1 to T2 lesion load with EDSS became stronger over time in the SPMS group.
Table 2. Spearman rank correlations between cross-sectional and longitudinal clinical and MRI data for the whole group and for relapsing-remitting (RRMS) and secondary progressive (SPMS) groups separately
Longitudinal comparisons: change in disability (EDSS) versus changes on MRI.
Significant correlations were found between the change in disability and the relative change from baseline in T1 and T2 lesion load, but exclusively so in the SPMS patients (Table 2). Again this relationship was stronger for T1 than for T2 lesion load (SRCC = 0.81 and SRCC = 0.66, respectively). Absolute changes in lesion load did not correlate with changes in disability in this patient sample. Figure 2 illustrates the relationship between relative T1 lesion load changes and progression rate for both RRMS and SPMS groups. The initial T1 lesion load correlated significantly with the absolute change in T1 per year (SRCC = 0.66, p < 0.0001). A significant correlation between the disability at the end of the study and the initial T1 lesion load was found only in the secondary progressive patients (SRCC = 0.65, p = 0.005).
Figure 2. Scatterplot of percent change in T1 lesion load versus change in EDSS per year (= progression rate) for the SPMS group (top panel) and the RRMS group (bottom panel). Regression curves shown as illustration.
Multiple regression analysis (forward and backward stepwise) was performed to determine the main factors influencing the change in EDSS per year during follow-up (dependent variable) in this sample of patients. Independent variables investigated were age, disease duration, EDSS at entry, progression rate before entry in the study (EDSS/disease duration), initial T1 and T2 lesion loads, absolute and relative changes in T1 and T2 lesion load, and the ratio of T1 to T2. This yielded a model with a multiple R value of 0.45 (R squared = 0.20). The following parameters were included in the model: progression rate before entry (p = 0.02) and absolute change in T1 lesion load (cm2 per year, p = 0.04). T2 imaging data were not included in this model and apparently provide no independent information.
T1 versus T2 imaging data.
When the changes in T1 and T2 imaging data between RRMS and SPMS patients are compared, some differences are noted. In the SPMS group, a correlation was found between the absolute change in T1 per year versus the absolute change in T2 per year, which was better than that found in the RRMS group (SRCC = 0.82, p < 0.0001; SRCC = 0.45, p = 0.014, respectively). This was also true in a comparison of the relative change from baseline for T1 and T2 (for SPMS, SRCC = 0.80, p < 0.001; for RRMS, SRCC = 0.34, p = 0.08; Figure 3). These correlations illustrate that in the SPMS group, changes on T2 are more frequently accompanied by changes on T1 than in the RRMS group. This is also reflected by the trend toward a significant difference between both groups in the ratio of T1 lesion load over T2 lesion load (see Table 1). Figure 4 illustrates the appearance of the hypointense lesions and the different balance between T1 lesion load and T2 lesion load in patients with high and low disability.
Figure 3. Relationship between T1 and T2 lesion load changes per year for SPMS (top panel) and RRMS (bottom panel). Regression curves shown as illustration, not as indicative of statistically significant correlation (see text for further details).
Figure 4. Two patients, one with SPMS (A, B), the other with RRMS (C, D). T1-weighted images are represented on the left side, T2-weighted images on the right side. Note that the lesion load on T2 is nearly identical, while T1 lesion load is clearly higher in the SPMS patient (example of a hypointense lesion is indicated by the arrow). The EDSS for the SPMS patient was 6 at the time of the scan; the EDSS of the RRMS patient was 3.
Reproducibility.
In a sample of five patients, the mean percent difference between a first and a repeat assessment of a set of scans was 5% for T1 lesion load and 3% for T2 lesion load. This is comparable with the intraobserver variation found in our previous study. [18]
Discussion.
To evaluate computer-aided quantification of hypointense lesions on T (1-weighted) MR images (black holes) as a possible means to solve the clinical-radiological paradox between MRI and standard clinical assessments, we correlated the MR findings with the concurrent clinical data in 46 patients with clinically definite MS at baseline and after a median follow-up of 40 months. The traditionally used MRI measure of disease burden is the T2 lesion load. In accordance with other studies, a low, marginally significant correlation was found between the EDSS at entry and the initial T2 lesion load for the whole group (Table 2). [7,8] This correlation was higher for T1 lesion load (Table 2), especially in SPMS patients. That the correlation is still not very strong might be inherent to the incomplete imaging of the CNS (the spinal cord and optic nerves were not imaged) and the dependence of the clinical rating scale used on spinal symptoms and signs.
Comparing the change over time in T1 lesion load and T2 lesion load with the change in disability (defined by EDSS), we found a strong correlation between the relative increase in T1 lesion load and the change in disability in the SPMS group; a weaker, but still significant, correlation was found for relative (percent) change in T2 lesion load in this group (Table 2). Figure 2 illustrates this discrepancy between RRMS and SPMS. The possible significance of this difference is discussed below. A second observation is that the absolute change in lesion load (cm2 per year, T1 or T2) was not significantly correlated with clinical changes in either group of patients. A possible explanation is that when the initial lesion load is low, small absolute changes give rise to large relative changes and vice versa. Relative changes could be more important than absolute ones in assessing burden of disease, because cranial MRI only samples part of the CNS. Perhaps the relative changes give more information about the global disease activity in the CNS system than do the absolute changes (in cm2) in brain lesion load only. Furthermore, some patients produce more lesions per year than others, and apparently, only changes in their individual rate of lesion formation predict changes in disability. Also, the variance of the absolute changes shows a skewed distribution, which hampers the establishment of a correlation; when disease duration or baseline lesion load are taken into account, this variance becomes more normal. This would explain why in the regression analysis the absolute changes became more important than relative changes, because other independents (e.g., progression rate at entry (defined as EDSS/disease duration) normalized the variance in absolute T1 changes.
The differential behavior of the lesions as seen on T1-and T2-weighted images is of interest. In the RRMS patients, many T2 changes are unaccompanied by T1 changes; in contrast, in SPMS patients, there is a more linear relationship between accumulation of hypointense lesions on T1-and hyperintense lesions on T2-weighted images (Figure 3). If hypointense lesions represent areas of severe demyelination, axonal loss, or gliosis, [13,14,16,21] this higher rate of new T2 lesions accompanied by signal loss on T1 could reflect a deficit in available repair mechanisms in secondary progressive patients. This is most likely not a qualitative but a quantitative phenomenon, since we also found, albeit to a lesser extent, hypointense lesions in RRMS patients. A number of studies (e.g., the recent study on copolymer 1), [22] have shown that RRMS patients with high EDSS scores have higher clinical relapse rates. This is also corroborated by an increase in gadolinium-enhanced lesion frequency in actively relapsing-remitting and early progressive patients. [5,23,24] This higher "inflammatory" activity perhaps exhausts repair mechanisms, leading to persistent structural loss without significant remyelination, in turn leading to persistent functional loss (failure of remission). [25]
Several studies have addressed the potential of magnetization transfer (MT) imaging in evaluating the extent of structural loss. [13,14,17] Gass et al. found a correlation of -0.44 between magnetization transfer ratio and EDSS, while the correlation of T2 lesion load and EDSS in that study was 0.33. [14] The magnitude of this improvement in correlation closely parallels the findings in our study using T1 lesion load and EDSS (Table 2). Recent data from our group [26] have shown that T1 signal intensity changes are closely related to MT ratio changes, and, as suggested by Hiehle et al., the hypointense lesions seen on standard MRI may indeed reflect areas of structural loss. [13] We therefore speculate that the accumulation of T1 lesions could be the MR equivalent of a failure of remission. As the T1-weighted images were obtained after administration of gadolinium-DTPA, some of the hypointense lesions could be missed, because they enhanced and became iso- or hyperintense with the white matter. This would relate to studies that found a lower MT ratio in some acute enhancing lesions. [13] Although the process has not been fully elucidated, it is conceivable that demyelination in these acute lesions is still reversible. Once the temporary nature of this phenomenon is established, not counting these lesions could actually be more accurate when comparing the hypointense lesion load with persistent disability.
Because T1-weighted SE images are easy to obtain, they offer significant advantages over the more sophisticated MR techniques such as MT imaging and MR spectroscopy, especially in multicenter studies. In such studies, the performance of these "simple" T1 SE sequences should nevertheless be strictly controlled, because of the observed incidental MT effects in multislice imaging that result in increasing loss of contrast between lesions and normalappearing white matter with increasing number of slices. [27]
Extrapolation of our findings to more heavily T1-weighted imaging sequences such as inversion recovery or inversion-prepared gradient-echo techniques should be made with care, since almost all hyperintense T2 lesions appear dark on T1 when these sequences are used, probably leading to the same high sensitivity-low specificity problem that hampers pathologic differentiation on standard T2-weighted SE imaging.
In conclusion, quantitative assessment of hypointense lesions resolves part of the clinical-radiological paradox encountered when correlating MRI lesion load with clinical disability in MS and should be further investigated as a possible surrogate marker of disease progression. [28] The observed differences between RRMS and SPMS patients probably relate to a quantitative difference in repair mechanisms. Studies confronting MR data collected with standard imaging, spectroscopy, and MT imaging with histologic data are necessary to further elucidate the nature of these hypointense lesions.
In view of these findings, we propose that the focus of interest in the long-term MRI follow-up of MS patients should be shifted from unidentified bright objects on T2 to "black holes" on T1.
Acknowledgment
The expert assistance of Ton Schweigman in performing the MRIs was greatly appreciated.
- Copyright 1996 by Advanstar Communications Inc.
REFERENCES
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Alert Me
Recommended articles
Correlating MRI and clinical disease activity in multiple sclerosis
A longitudinal study of cerebral glucose metabolism, MRI, and disability in patients with MS
TNFα production by CD4+ T cells predicts long-term increase in lesion load on MRI in MS
Assessing information in T2-weighted MRI scans from secondary progressive MS patients